#50 Autostart settings don't apply until reboot on fresh Fedora 34 installations
Closed: Can't fix 2 years ago by ngompa. Opened 3 years ago by eiglow.

When upgrading from an older version of Fedora, Autostart seems to work fine, I use it on my desktop. But from what I have heard from one user, and my own testing in a VM, Autostart seems to not work with a fresh f34 install.

I tested this by telling it to automatically start Dolphin, logging out of the session, and then logging back in. I couldn't reproduce this issue on my KDE Neon Unstable VM, which is the main reason I'm reporting it here.

I looked through journalctl /usr/bin/plasmashell, and spotted the following line:

kf.kio.gui: Failed to register new cgroup: "app-autostart-4e12cde8171d4a88b8c761993c183409.scope" "org.freedesktop.DBus.Error.AccessDenied" "Failed to get file context on /home/alex/autostart.desktop."

After fully rebooting my VM, Dolphin seems to autostart properly though.

At the time of filing, I installed the 34 KDE spin beta on a physical machine, and (repeated) reboots did not help. The .desktop file was created correctly though.

This should be reported upstream, bugs.kde.org, pretty sure this should be fixable

On a current openSuse Tumbleweed I did not have this problem. That's why I thought it was Fedora specific.

It's not Fedora-specific, it's likely related the usage of systemd user session based startup, which is still upstream functionality, just something we're using first.

https://bugs.kde.org/show_bug.cgi?id=435947 I have created an upstream bug report, please let me know if I wrote anything wrong

According to Nate, this should be fixed in systemd 249.

We're on systemd 248, however, it might be possible to backport the fix to our systemd package. @zbyszek, do you know if this can be fixed for us?

kf.kio.gui: Failed to register new cgroup: "app-autostart-4e12cde8171d4a88b8c761993c183409.scope" "org.freedesktop.DBus.Error.AccessDenied" "Failed to get file context on /home/alex/autostart.desktop."

That message seems to be systemd when it tails to call getfilecon_raw().

We're on systemd 248, however, it might be possible to backport the fix to our systemd package.

This patch is a bit large: we usually don't backport new features. But it's not too big, and if it really fixes the issue, this can be done.

But are you sure that this patch is what you need? The only error message that was quoted above is about selinux...

The problem persists in the final release of Fedora 34: setting items to autostart in Plasma has no effect whatsoever.

Is anything being done about this?

Upstream issue has an active pull request under review

Need them to sort it out first as a prerequisite

Thanks for your update @rdieter ! I suppose it's impossible to say, but I'll ask anyway: is there any ETA of this landing in Fedora?

Once fix lands upstream, we can pull it into fedora. As to when the former happens, you'd have to ask them (systemd upstream devs working the issue).

Ah, I do see the pull request was recently merged, I'll poke systemd maintainers in
to see if we can get that fix into fedora.

Metadata Update from @ngompa:
- Issue assigned to rdieter
- Issue set to the milestone: Fedora 34

2 years ago

As this is essentially (still) an upstream systemd that is farily well-understood at this point, I think we can probably consider the issue resolved as far as the kde-sig is concerned (fixing this is out of our hands).

I'm not sure I agree. The bug is present, identified, fixed upstream, but not yet merged into Fedora 34. That makes it an active bug.

Active bug yes, but is there anything left for the kde-sig to actually do about it?

I'd say that conditions for closing the issue are:

  1. The fix is merged and pushed to users.
  2. Fedora 33 goes out of support.

It's useful to know the issue persists in my view.

That's what bugzilla is for. I hope you're not suggesting we track/document every unresolved bugzilla bug here in pagure?

The primary purpose of the fedora-kde issue tracker is to track issues that the kde-sig are (or ought to be) working on. In this case, the work on our side is done.

I did not propose that=2E

Then why treat this unresolved bug any differently?

From the KDE SIG side, we can't do anything. If you want something to happen, go to the BZ instead: https://pagure.io/fedora-kde/SIG/issue/50

Metadata Update from @ngompa:
- Issue close_status updated to: Can't fix
- Issue status updated to: Closed (was: Open)

2 years ago

Login to comment on this ticket.